Automatic Facial Age Estimate Based on Convolution Neural Network

Jiancheng Zou, Xuan Yang, Honggen Zhang, Xiaoguang Chen

Abstract: As a new biometric recognition technology, automatic age estimation based on facial image has become an important subject in the field of computer vision and human-computer interaction (HCI). And automatic facial age estimate system has been increasingly used in criminal investigation, image retrieval and intelligent monitoring in recent years. Therefore, the research of facial age estimation has a broad prospect. Convolution neural network (CNN) as a deep learning architecture can extract the essential features of the facial image with a better effect than traditional methods, especially in the case of large changes in imaging shooting conditions. In this paper, an improved method of facial age estimate based on CNN is proposed. By considering the number limitation of the existing age estimation data sets, we adopt the method of fine-tuning the existed network model. The recognition rate can be increased by 3% based on the proposed method. A facial age estimate system has been constructed for applications and the experimental results show that the system can meet the real-time application needs.